Can some please tell me how “rasa test nlu” picks the test data, how it ensures that same data which is used in training is used in test again ?
How to pass a custom set of test data for rasa test nlu so that I can benchmark rasa nlu accuracy with some custom nlu component accuracy ?
Thanks Shakti
Hi @shaktisd, have you looked at this part of our blog? Evaluating Models
If you don’t provide a test set, it will test on your training data. You can use cross-validation to mitigate that effect if you want real testing.
If you have your own test set, you can pass it via -u test_set.md, and you will have to pass the trained model with the -m flag.
Hello! Is there a way to tun “rasa test” using the REST API?
My team has built an application for managing Rasa training data and training actions, and it would be nice to perform training model evaluation using it!